HAL Id: hal-02973353
https://hal.archives-ouvertes.fr/hal-02973353
Submitted on 21 Oct 2020
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Distributed under a Creative Commons Attribution| 4.0 International License
mutations in the damage response gene HELB
Marina Naval-Sánchez, Laercio R. Porto-Neto, Diercles F. Cardoso, Ben J.
Hayes, Hans D. Daetwyler, James Kijas, Antonio Reverter
To cite this version:
Marina Naval-Sánchez, Laercio R. Porto-Neto, Diercles F. Cardoso, Ben J. Hayes, Hans D. Daetwyler,
et al.. Selection signatures in tropical cattle are enriched for promoter and coding regions and reveal
missense mutations in the damage response gene HELB. Genetics Selection Evolution, BioMed Central,
2020, 52 (1), pp.27. �10.1186/s12711-020-00546-6�. �hal-02973353�
RESEARCH ARTICLE
Selection signatures in tropical cattle are enriched for promoter and coding regions
and reveal missense mutations in the damage response gene HELB
Marina Naval‑Sánchez
1,6*, Laercio R. Porto‑Neto
1, Diercles F. Cardoso
1,2,7, Ben J. Hayes
3, Hans D. Daetwyler
4,5, James Kijas
1and Antonio Reverter
1Abstract
Background: Distinct domestication events, adaptation to different climatic zones, and divergent selection in pro‑
ductive traits have shaped the genomic differences between taurine and indicine cattle. In this study, we assessed the impact of artificial selection and environmental adaptation by comparing whole‑genome sequences from European taurine and Asian indicine breeds and from African cattle. Next, we studied the impact of divergent selection by exploiting predicted and experimental functional annotation of the bovine genome.
Results: We identified selective sweeps in beef cattle taurine and indicine populations, including a 430‑kb selective sweep on indicine cattle chromosome 5 that is located between 47,670,001 and 48,100,000 bp and spans five genes, i.e. HELB, IRAK3, ENSBTAG00000026993, GRIP1 and part of HMGA2. Regions under selection in indicine cattle display significant enrichment for promoters and coding genes. At the nucleotide level, sites that show a strong divergence in allele frequency between European taurine and Asian indicine are enriched for the same functional categories. We identified nine single nucleotide polymorphisms (SNPs) in coding regions that are fixed for different alleles between subspecies, eight of which were located within the DNA helicase B (HELB) gene. By mining information from the 1000 Bull Genomes Project, we found that HELB carries mutations that are specific to indicine cattle but also found in tau‑
rine cattle, which are known to have been subject to indicine introgression from breeds, such as N’Dama, Anatolian Red, Marchigiana, Chianina, and Piedmontese. Based on in‑house genome sequences, we proved that mutations in HELB segregate independently of the copy number variation HMGA2‑CNV, which is located in the same region.
Conclusions: Major genomic sequence differences between Bos taurus and Bos indicus are enriched for promoter and coding regions. We identified a 430‑kb selective sweep in Asian indicine cattle located on chromosome 5, which carries SNPs that are fixed in indicine populations and located in the coding sequences of the HELB gene. HELB is involved in the response to DNA damage including exposure to ultra‑violet light and is associated with reproductive traits and yearling weight in tropical cattle. Thus, HELB likely contributed to the adaptation of tropical cattle to their harsh environment.
© The Author(s) 2020. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/publi cdoma in/
zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Background
The domestication of wild aurochs (Bos primigenous) in two distinct locations, in the Middle East (~ 10,000 years ago) and the Indian subcontinent (~ 8000), resulted in the separate evolution of two cattle lineages and in
Open Access
*Correspondence: m.navalsanchez@imb.uq.edu.au
1 CSIRO Agriculture & Food, 306 Carmody Rd., St. Lucia, Brisbane, QLD 4067, Australia
Full list of author information is available at the end of the article
divergences between the genomes of taurine (Bos primig- enous taurus) and indicine (Bos primigenous indicus) cattle. In general, they occupy distinct geographic and climatic locations worldwide [1, 2]. Taurine cattle are mostly found in temperate environments, whereas indi- cine breeds are highly adapted to environments with con- stant high temperatures [3]. Besides adaptation to heat, other environmental adaptation traits such as disease and parasite resistance, and differences in human herd man- agement and selection processes have driven different patterns of genomic variation between these cattle sub- species. This offers the opportunity to identify genes that are involved in adaptation, within the genetic context of a single species. The identification of genomic regions impacted both by human selection and climate adapta- tion will help understand how changes at the genome level modulate changes in phenotype, which holds high promises to improve animal breeding processes for pro- duction, health, and welfare [4]. Previous analyses using single nucleotide polymorphism (SNP) arrays [5–13] and whole-genome sequences [14, 15] have identified candi- date regions and potential genes under selection in vari- ous cattle breeds. However, compared to other domestic species for which selection is known to impact mostly conserved elements, transcription start sites or regula- tory regions [16–18], to our knowledge, there has been no effort to understand the impact of selection at the functional-genomic level in cattle. To date, the lack of functional genomic information has limited the attempts to analyze the impact of selection and evolutionary diver- gence in cattle and in other livestock species.
Recently, an international effort entitled ‘The Func- tional Annotation of Animal Genomes’ (FAANG; https ://www.anima lgeno me.org/commu nity/FAANG /index ) has been launched and aims at addressing the above issue by experimentally identifying regulatory regions in the genomes of many tissues and at several stages of develop- ment [19]. Meanwhile, our group has recently provided a first draft of cattle and sheep functional regulatory regions based on the identification of orthologous regula- tory regions [18, 20] in other species from the human and mouse ENCODE [21, 22] and RoadMap consortia [23].
In this work, our objectives were to investigate genomic differences between Bos taurus and Bos indicus in the European versus Indian subcontinents and between Afri- can taurine and indicine breeds, to identify candidate selective sweeps in these populations, and assess their enrichment for distinct functional elements.
Methods Samples
We retrieved 440 whole-genome sequences from the 1000 Bull Genomes Project (Run6, Bos taurus, and Bos
indicus) for the 18 breeds that were chosen to consti- tute the reference population for imputation (Table 1) [24, 25]. The dataset contained 186 European tau- rine, 102 Asiatic indicine and 80 crossbred genomes as well as a subset of African samples from 12 taurine, 41 Sanga (ancient stabilized taurine × indicine crossbred [26], and 19 indicine individuals (Table 1). These breeds were selected to capture the lineages that are relevant to the beef industry since most tropical beef cattle are a genomic mosaic of indicine, African Sanga, European and African taurine cattle [27–29]. Thus, no dairy breeds were included in the study. Breeds were grouped accord- ing to their phenotypes and to known genomic crosses, i.e. taurine (humpless), indicine (with hump), admixed or African Sanga, the two latter being stabilized composite breeds [27, 30–33]. The selected animals were sequenced on an Illumina HISeq sequencer at an average coverage of 11.68 that ranged from 1.84 to 44.17.
Mapping, variant detection and imputation
The selected genome sequences were processed through the 1000 Bull Genomes Project pipeline [34]. Before sequence alignment, data were trimmed for adaptor sequences using Trimmomatic [35] and reads with a Phred quality score lower than 20, or with a read length shorter than 50% of the standard length were discarded.
The genome sequences were aligned to the UMD3.1 reference genome [36] with the BWA-MEM algorithm, using default parameters [37]. Duplicates were removed using Picard’s MarkDuplicates tools (http://broad insti
Table 1 Whole-genome sequences used in the study Name Animal source Sample size Genome of origin
Brahman Australia 90 B. indicus
Nelore Brazil 5 B. indicus
Gir Brazil 6 B. indicus
Shahiwal India/Pakistan 1 B. indicus
Composite Australia 56 Indicine‑taurine
Brangus USA 5 Indicine‑taurine
Santa Gertrudis USA 4 Indicine‑taurine
BeefMaster USA 15 Indicine‑taurine
Charolais France 128 B. taurus
Angus Great Britain 51 B. taurus
Shorthorn Great Britain 5 B. taurus
British shorthorn Great Britain 2 B. taurus
N’ Dama Africa 12 B. taurus
Uganda‑mix Africa 26 Sanga
Africander Africa 5 Sanga
Ankole Africa 10 Sanga
Ogaden Africa 9 B. indicus
Boran Africa 10 B. indicus
tute.githu b.io/picar d/) and local realignment of the reads around InDels was done with the GATK [38] tool Indel- Realigner. Variant calling was performed by applying the GATK tool Best Practises [38]. All raw variants were called with the GATK [38] tool HaplotypeCaller based on the Bos taurus reference genome UMD3.1 and all raw variant VCF files were combined via the Genotype GVCF tools to produce a single VCF file. Genetic variants from the sequenced animals were extracted and filtered to retain only bi-allelic variants that had at least four copies of the minor allele. Sequences of the filtered variants were phased and imputed with the Eagle [39] and FImpute 2.2 [40] software, respectively. The analysis resulted in the detection of 39,679,303 high-quality SNPs, of which 24,080,747 were considered common SNPs (minor allele frequency (MAF) ≥ 0.05). Genetic diversity estimates were obtained by using PLINK v1.9 and PCA (https ://
www.cog-genom ics.org/plink 2) [41]. The VCFtool v.0.16 (–het) was used to calculate the observed homozygosity and heterozygosity as well as the inbreeding coefficient, F, for each individual [42]. Individual heterozygosity and F-values were plotted per breed and genome of origin using R version 3.5.2.
Selective sweeps
Allele frequency differences between taurine and indi- cine populations were measured using the FST index (Weir and Cockerham method [43]). Average FST values were plotted in 20-kb overlapping genomic bins (with a number of SNPs > 10) with a 10-kb step-size. Nucleotide diversity (π) was measured in each population within the same 20-kb genomic bins. The ratio of indicine to taurine π was used to identify differences in nucleotide diver- gence between populations. The combined analysis of FST
and π ratio (indicine/taurine) was used to identify can- didate sweeps. The Z-transformed product of FST and π ratio values was declared significant if the genome-wide threshold was higher than 5.08, which represents a Bon- ferroni adjusted p-value lower than 0.05.
Biological and phenotypical enrichment analysis
We performed a locus-based gene ontology enrichment with the GREAT v.3.0.0 software package [44]. Candidate selective sweeps (bins and/or regions) were translated to human coordinates (GRC37/hg19) using the liftOver tool (minMatch = 0.1) [45]. GREAT associates regions to genes and then performs a binomial (gene) and a hyper- geometric test (region) to calculate the enrichment for biological terms, processes, Mouse Genome Informat- ics (MGI) database phenotypes, and Human Phenotype Ontology from OMIM. The default option ‘Basal plus extension’ association rule assigns genomic regions with genes, i.e. each gene is associated to a basal regulatory
domain that extends 5 kb upstream and 1 kb down- stream of the transcription start site (TSS) (regardless of the other nearby genes). In addition, each gene has an extended regulatory domain in both directions up to 1000 kb or until the basal domain of the nearest gene.
Functional annotation of the cattle genome
We used the UMD3.1. version 1.87 assembly of the bovine genome and derived the following functional annotation tracks:
• Gene: gene coordinates expanding the exonic and intronic regions of a gene.
• CDS: coding sequences coordinates within a protein- coding gene.
• Intron: intronic coordinates were calculated as gene coordinates minus the CDS regions.
• Intergenic regions: whole-genome regions absent of gene coordinates annotation.
• 1-kb upstream: 1-kb regions upstream of the tran- scription start site (TSS) of the annotated protein- coding gene.
• 1-kb downstream: 1-kb regions downstream of the transcription end site (TES) of the annotated pro- tein-coding gene.
• UTR: 3′ and 5′ UTR regions.
Next, we used predicted regulatory elements from our previous study [20] in which human regulatory elements from three distinct human regulatory databases, i.e.
ENCODE, FANTOM and Epigenomics Roadmap, were projected onto cattle coordinates by reciprocal liftOver (minMatch = 0.1) [45]. The original or full set was fur- ther processed by applying different filters and thresholds including those for expression in bovine tissues [20]. The following datasets were included in the study:
• Human Projection All dataset: all predicted regula- tory elements, proximal (promoters) and distal reg- ulatory elements projected onto the bovine genome from three human databases, ENCODE, FANTOM, Epigenomics Roadmap. No filtering.
• Human Projection Proximal Elements: all proximal (promoter) regulatory elements from the same three databases.
• Promoter: FANTOM5 promoter atlas that was gen- erated experimentally with CAGE data from almost 1000 tissues and cell lines [46] and projected onto cattle coordinates [20]. CAGE is a methodology for the detection of core promoter regions that bind the transcriptional machinery [47].
• Human Projection EnhG: all genic enhancers (EnhG) regions from the Epigenomics Roadmap
database [23]. EnhG are enriched for H3K4me1 and H3K36me3 chromatin marks and correspond to enhancers that overlap with exonic regions [23].
• Human Projection EnhBiv: enhancer bivalent (Enh- Biv) regions from the Epigenomics RoadMap data- base [23]. EnhBiv are associated with H3K4me1 and H3K27me3 chromatin marks [23].
• Human Projection Enh: enhancers (Enh) regions that are detected in the RoadMap Epigenomics database.
Such enhancers are associated with H3K4me1 chro- matin marks and tend to be distal regulatory ele- ments [23].
• Human Projection Proximal transcription fac- tor binding sites (TFBS): proximal TFBS from the ENCODE dataset [21].
• Human Projection Distal TFBS: distal TFBS from the ENCODE dataset [21].
• Human Projection Filtered set: whole dataset pro- jected onto cattle coordinates after filtering.
Finally, we exploited publicly available experimen- tal epigenomic marks present in the cattle genome, including:
• ATAC-seq cattle FR-AgENCODE data: the assay for transposase accessible chromatin (ATAC) identi- fies nucleosome-depleted regions in the genome, which are enriched for regulatory functions. The FR-AgENCODE pilot study performed ATAC-seq in CD4+ and CD8+ cells (http://www.frage ncode .org/
resul ts.html) [48].
• Experimental chromatin marks in the liver obtained from a comparative analysis across 20 mammalian species [49], i.e. the ArrayExpress database with accession number E-MTAB-2633).
• Cattle H3K4me3: genomic coordinates that are sig- nificantly enriched for H3K4me3 chromatin marks in the Bos taurus liver. H3K4me3 is associated with promoter regions.
• Cattle H3K27ac: genomic coordinates that are signif- icantly enriched for H3K27ac chromatin marks in the Bos taurus liver. H3K27ac is associated with active regulatory function.
• Cattle H3K27ac only: genomic regions that are sig- nificantly enriched for H3k27ac chromatin marks but with no enrichment for H3K4me3 chromatin marks in the Bos taurus liver.
Assessment of the functional enrichment of selective sweeps
To assess the enrichment of genomic region sets, i.e.
selection sweeps, for various functionally annotated
genomic elements within the cattle genome, we used the R/Bioconductor package locus overlap analysis (LOLA) [50]. This tool requires (i) a ‘query set’, which is the list of genomic regions to be tested for enrichment.; (ii) a
‘reference set’ or a list of genomic regions to be tested for overlap with the ‘query set’; and (iii) a ‘universe set’, which is a background set of regions that could have been included in the query set. LOLA performs a Fisher’s exact test with a false discovery rate correction to assess the significance of the overlap in each pairwise comparison between the ‘query set’ and each entry in the ‘reference set’ [49]. We investigated the enrichment of detected candidate sweeps (query set) for a collection of distinct cattle functional elements (reference set). These include annotations (i) that are derived from the reference assem- bly UMD3.1 v.1.87; (ii) on predicted regulatory elements in the cattle genome based on the translation of human epigenomic marks coordinates from ENCODE and RoadMap epigenomics [20]; and (iii) on experimentally available epigenetic marks for the cattle genome [49] and Fr-AgENCODE ATAC-seq datasets [48]. The universal set was defined as a list of 20-kb genome-wide bins that were inputted in the selection sweep analysis and could potentially be found under selection.
Analysis of divergent allele frequencies
between populations and across functional categories To assess whether divergent SNPs between populations were enriched for certain functional categories, we esti- mated the reference allele frequency (RAF) per SNP and per population using VCFtools (–freq) [42]. Then, we cal- culated their absolute allele frequency difference between populations ΔAF = abs (AFtaurine − AFindicine). Next, we binned SNPs by ΔAF in steps of 0.1 i.e. (ΔAF = 0.00–0.10, 0.01–0.20, etc. up to 0.90–1.00) resulting in 10 bins. These bins were intersected with functional categories i.e. cod- ing exons, intronic, intergenic regions, etc., as described in the Cattle functional annotation methods section. For the ΔAF bins, the proportions of SNPs in each functional category were determined by using the software bedtools intersect [51]. M-values (log2-fold change) of the rela- tive frequencies of SNPs in each functional category were calculated by comparing the frequency of SNPs per func- tional category in a specific bin with the corresponding frequency across all bins (expected value). Statistical sig- nificance of the deviation from the expected values was assessed using a Chi squared test. It should be noted that the FST at the SNP level, which is a measure of population divergence that accounts for ΔAF across populations, and the variance of the allele frequency within each popula- tion, could have been used for the analysis and binning.
In our study, since both metrics were highly correlated (r2 = 0.975, we chose ΔAF, which is easier to use [17, 52]).
HELB allele frequency across 2707 animals from the 1000 Bull Genomes project
Data processing and variant calling for the 1000 Bull Genomes sequences are described in [25]. Run6 of the project (released on March 2017) included 2707 ani- mals from 97 breed groups, with 2379 animals classi- fied as taurus and the remainder as unknown, indicus, or admixed [25]. We calculated allele frequencies per breed based on the breed classification provided by the 1000 Bull Genomes project.
HMGA2‑CNVR
We exploited a collection of in-house whole-genome sequence data from commercial breeding animals including five Africander, 56 tropical composites and 10 Brahman. All these animals are part of the 1000 Bull Genomes Project data collection. DNA was extracted from either blood or semen samples from each animal following a standard protocol. Paired-end short insert libraries were sequenced on the Illumina HiSeq 2000 platform. Reads were mapped against the cattle reference assembly UMD3.1/bosTau6 [36] using the BWA aligner v0.7.1 (bwa mem, default param- eters) [37]. Duplicates reads were marked using Pic- ard tools (http://broad insti tute.githu b.io/picar d/). We assessed the existence of copy number variants (CNV) in the known HMGA2-CNV region on chromosome 5 between 48,074,233 and 48,080,443 bp (~ 6.2 kb) [53]
by comparing the coverage in the CNVR versus the coverage along the whole chromosome 5. In addition, the alignments of all 71 animals were visualized with Integrative Genomics Viewer (IGV) [54] to confirm the existence of reads that harbor a duplication of the HMGA2-CNV.
Results
Genetic variation between taurine, indicine and admixed cattle
To evaluate the genomic relationships between samples, we performed a principal component analysis (PCA) across all samples and datasets (see Additional file 1: Fig- ure S1). In agreement with previous reports [55–57], PC1 (84.02% of variability) captured the taurine/indicine ori- gin and PC2 (11.60% of variability) captured the African origin of the samples. The same PCA without the African samples resulted in PC1 capturing the taurine/indicine origin (77.74% of the variability) and PC2 (7.92% of the variability) dividing the taurine breeds along the Angus Charolais axis (Fig. 1a) and (see Additional file 1: Fig- ure S2). Since the African samples represented a much smaller dataset (Table 1), we report the comparison of the African taurine versus indicine cattle, separately.
European and Asiatic breed variant calling resulted in 38,865,098 high-quality SNPs, of which 16,918,921 were shared among the three indicine, admixed and taurine populations (Fig. 1b). The number of private SNPs was larger for the indicine breeds (2,712,827) than for the taurine (1,690,752) or the admixed (500,723) breeds (Fig. 1b). The nucleotide diversity (π) and heterozy- gosity (het) values were higher for the indicine breeds (π = 0.32% and het = 0.20) than for the admixed and taurine breeds (π = 0.27% and 0.15% and het = 0.18 and 0.09, respectively) as shown in Fig. 1c and Additional file 1: Figure S3. The coefficient of inbreeding F was lower for indicine (− 0.07) and admixed (0.05) than for the taurine breeds (0.50), as shown in Additional file 1: Figure S4. The same trend was observed for the samples from Africa (indicine cattle: 1,767,058 private SNPs, π = 0.32%, het = 0.21, F = − 0.11; Sanga cattle:
3,794,711 private SNPs; π = 0.29%, het = 0.19, F = 0.006;
taurine N’Dama: 510.325 private SNPs; π = 0.17%, het = 0.12, F = 0.34; (see Additional file 1: Figures S3–
S5). We observed a smaller number of private SNPs and a lower nucleotide diversity for the taurine than the indicine breeds, which agrees with the more intense artificial selection of their production systems and with the evidence of taurine introgression in indicine cattle [56, 58, 59].
Finally, we observed that although most SNPs were common between indicine and taurine breeds, the cor- relation of the reference allele frequency (RAF) bins between these two breeds was low (R2 = 0.41), compared to that between indicine and admixed (R2 = 0.78) and between taurine and admixed (R2 = 0.84) breeds (Fig. 1d–
f). Similar results were found for African cattle (see Addi- tional file 1: Figure S5). This indicates that the genomic divergence between indicine and taurine breeds is higher than between domestic sheep (Ovis aries) and their wild counterpart (Mouflon Ovis orientalis) (R2 = 0.79) [18].
Genomic regions under selection in European taurine and Asian indicine cattle
By pooling genomes into groups of subspecies and com- paring the patterns of variability, we sought to identify genomic regions and genes that are putatively involved in their phenotypic and behavioural differences. The FST (see Additional file 1: Figure S6) and the ratio of indicine to taurine π were plotted for each 20-kb genomic bin (Fig. 2a, b) and (see Additional file 2: Table S1), revealing 657 candidate bins under selection in the taurine genome and 242 in the indicine genome (P-adj < 0.05) (see Addi- tional file 3: Table S2 and Additional file 4: Table S3). Bins that were closer than 50 kb apart were merged, which yielded 376 and 72 candidate selective sweep regions, for the taurine and indicine genome, respectively (average
sizes of 29,5 kb, and 52,5 kb, respectively) (see Additional file 5: Table S4 and Additional file 6: Table S5).
Inspection of the gene content in the taurine selective sweep regions revealed several genes that are known in cattle or other species. For example, for the melano- cortin 1 receptor (MC1R) gene, we found low π and high FST in taurine cattle, which is consistent with val- ues reported in the literature for taurine cattle, horses and pigs in studies on coat pigmentation patterns [8, 60, 61]. The leucine-rich repeats and immunoglobulin- like domains protein 3 (LRIG3) gene is known to be under selection in Charolais cattle (a predominant tau- rine breed in our study) and has been associated with elongated body axis [8]. Another gene of interest is the myosin 1A (MYO1A) gene that is known to be under divergent selection between taurine and indicine breeds and to influence pigmentation [61, 62].
Few outlier regions and genes were detected in the genomes of indicine cattle, but within those regions, our results confirmed several previously reported genes, such as LEM domain-containing protein 3 (LEMD3) on Bos taurus (BTA) chromosome 5 (BTA5) [8]. A major finding was a large selective sweep that spans 430 kb on BTA5 (47,670,001–48,100,000 bp) (Fig. 2c). This is the largest region under selection, which also displays the largest difference in π between indicine and taurine cat- tle (Fig. 2b, c) and (see Additional file 6: Table S5). This region is near fixation in indicine cattle and spans several genes including HELB, IRAK3, ENSBTAG00000026993, GRIP1, and part of HMGA2 (Fig. 2c). Published genome- wide association studies (GWAS) in tropical cattle, associated this region with traits including sheath score and yearling weight [63] or reproductive traits in tropi- cal cattle [64–66]. Finally, within this candidate indi- cine selective sweep, a tandem duplication of ~ 6.2 kb
0.000 0.005 0.010 0.015
0100200300400500600700
Taurine Admixed Indicine
0.0000.0020.0040.006
b c
d e f
Indicine
Taurine
Indicine
Admixed Admixed
Taurine
R2 = 0.41 R2 = 0.78 R2 = 0.84
0.25 0.50 0.75 1.00
0.25 0.50 0.75 1.00
0.25 0.50 0.75 1.00
0.25 0.50 0.75
1.00 0.25 0.50 0.75
1.00 0.25 0.50 0.75 1.00
log10(number SNPs)
6 4 2
a
-0.05 0.00 PC1 0.05
0.10
0.05
0.00
-0.05
PC2
Indicine
33.579.194 Admixed
33.682.893
Taurine 22.491.728
16.918.921 2.712.827
13.159.820
787.626 3.094.429 500.723
1.690.752
Frequency
Taurine
Admixed
Indicine Angus
Charelois
Fig. 1 Genetic variation and divergence. a PCA of the genetic distance was performed to assess the clustering of sequences according to their genome of origin, taurine (blue), indicine (orange) or admixed (green). For PCA breed composition, (see Additional file 1: Figure S2). b Proportion and number of private and shared SNPs for the taurine (blue), indicine (orange) or admixed (green) cattle datasets. c Estimated nucleotide diversity in 20‑kb genomic bins for the taurine (blue), indicine (orange) or admixed (green) sequences. d–f Correlations between estimated reference allele frequencies (RAF) between taurine (blue) and indicine (orange) (d), admixed (green) and indicine (orange) (e) and admixed (green) and taurine (blue) (f). An increasing number of SNP counts is related to warmer colours
(48,074,233–48,080,443 bp) was reported to affect the third and fourth introns of HMGA2 in Nellore cattle and to be associated with navel length (similar to sheath score) at yearling (Fig. 2c) [53]. Taken together, these results indicate this 430-kb selective sweep is relevant for selective breeding programs aimed at improving adapta- tion of cattle to tropical conditions.
Genomic regions under selection in African cattle
Analysis of African whole-genome sequences (see Addi- tional file 1: Figure S7 and Additional file 7: Table S6) resulted in the detection of 1194 20-kb bins for African taurine cattle (N’dama n = 12) and 324 in African Bos indicus (Boran n = 10, Ogaden n = 10) (see Additional file 8: Table S7 and Additional file 9: Table S8. After merging the 20-kb bins less than 50 kb apart, we defined 611 and 117 genome-wide regions under selection in African taurine and indicine breeds, with an average size of 35.5 and 42.1 kb, respectively) (see Additional file 10:
Table S9 and Additional file 11: Table S10).
African taurine cattle (N’Dama) inhabit regions that are infested with tsetse fly, and thus, have evolved mech- anisms to tolerate trypanosoma infection, including resistance to anaemia, its major clinical sign [67]. Our analysis captured genes that are associated with resist- ance to anaemia, under selection in African taurine cat- tle, and potentially related to trypanotolerance in cattle (see Additional file 10: Table S9). These include, erytro- cyte membrane protein brain 4.1 (EPB41), which encodes proteins of the red cell membrane skeleton and is asso- ciated with hematologic disorders in humans related with variable degrees of anaemia [68], and ferroportin (SLC40A1), a gene that is relevant for iron homeostasis [69] and was previously reported to be under selection in African taurine cattle [14].
Biological processes and phenotypes associated with candidate selective sweeps
Analysis of the taurine populations, revealed only two significantly enriched biological processes terms: regula- tion of catenin import to the nucleus (binomial test FDR q-value 1.46 × 10−3, hypergeometric test FDR q-value 1.92 × 10−2) and embryonic skeletal joint development (Binomial test FDR q-value 3.38 × 10−2, hypergeometric
FDR q-value 3.65 × 10−2). At the phenotype level, we found a significant enrichment of candidate selec- tive sweeps for mouse behavioural traits (Table 2) and (see Additional file 12: Table S11) and the top term was
“increased exploration in new environment” [binomial test p-value 2.08 × 10−14, Table 2 and (see Additional file 12: Table S11)], which is consistent with the reported behavioural differences between taurine and indicine cat- tle [70–74]. The other enriched terms for mouse pheno- types in regions under selection in taurine cattle relate to changes in pigmentation such as belly spot or hypo- pigmentation (see Additional file 12: Table S11). These results are consistent with the objectives of artificial selection for colour patterns in many species including cattle [75], pig [76, 77], horse [78, 79], and sheep [80]. In contrast, the only enrichment associated with indicine candidate sweeps was for human phenotypes related to body height (binomial test FDR q-value = 8.9 × 10−17, hypergeometric test FDR q-value = 2.91 × 10−02, GREAT v 1.8 Human Phenotypes), which involves genes such as HMGA2, KDM6A, LEMD3, FERMT1 (see Additional file 13: Table S12). In cattle, body weight is a trait that has been subject to various selection pressures over time and across breeds [81–83].
No functional significant term was enriched in African cattle selective sweeps as previously reported [14].
Functional annotation associated with candidate selective sweeps
Selection can differ depending on distinct genomic functional elements, such as coding elements or regu- latory elements, which are mostly related to changes in gene expression. To tackle this issue, we investigated the enrichment of previously detected candidate sweeps for a collection of experimental and predicted cattle func- tional elements (Fig. 3a) and (see Additional file 1: Figure S8, Additional file 14: Table S13 and Additional file 15:
Table S14). No functional enrichment was observed in taurine candidate sweeps (n = 357) (see Additional file 1:
Figure S8 and Additional file 14: Table S13). However, regions under selection in the indicine cattle (n = 72) pre- sented a significant enrichment for proximal and genic features (Fig. 3a) and (Additional file 15: Table S14). This Fig. 2 Candidate selective sweeps in taurine and indicine cattle. a Population differentiation index (FST) and relative nucleotide diversity between taurine and indicine cattle in genome‑wide 20‑kb genomic bins. Outlier bins that show evidence of selection in taurine breeds (blue) and indicine breeds (red). b Genome‑wide distribution of relative nucleotide diversity. Positive and negative values represent candidate sweeps in taurine and indicine cattle, respectively. Outliner bins are coloured in red. c IGV screenshot of chr5: 47,526,093–48,203,280. In red, the 430 kb long selective sweep in Asian indicine cattle: spanning GRIP1, HELB, IRAK3, ENSBTAG00000026993, LLPH, and part of HMGA2. In green, a selective sweep in GRIP1 in European taurine cattle. In blue, the 6.2 kb tandem duplication HMGA2‑CNVR reported by [53]. Below 10 variant files in vcf format for 10 Brahman animals, 10 Angus, and 10 Charolais
(See figure on next page.)
HELB; IRAK3;HMGA2
SRY FRY ENSB
TAG00000 01233
9
EFCAB6 TMTC 3
DOCK 8 GRIP
1
4 ln(π Indicine /π Taurine)
ln(π Indicine /π Taurine)
Fst
Fst
TRANK1, DCLK3,ITGA9
a
b
ACTR 2
MC1R SCAPER
ADA ENSB M18
TAG0000 001380
9
Chromosome
ln( Indicine / Taurine)
0 0 2
-2 -1 1 2 3
0.0 0.2 0.4 0.6 0.8
4 0.0
0.3 0.6
-4 -2 0 2 4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
HELB coding mutations TADS
Selective sweeps Indicine Selective 20Kb bins Indicine Selective Sweep Taurine HMGA2-CNVR
Brahman
Angus
Charolais
Gene
chr5: 47526093 - 48203280
GRIP1
U1
GRIP1 HELB
HELB FAU
IRAK3
LLPH ENSBTAT00000008243
5S_rRNA
HMGA2
bta-mir-763
c
is indicated by the enrichment for experimentally defined promoters that were identified by H3K4me3 analysis in bovine liver tissue [49], and for 1-kb upstream genic and intronic regions from the current UMD3.1 v.187 (Fig. 3a) and (Additional file 15: Table S14). Analysis of African cattle selective sweeps also showed a significant enrich- ment for proximal features including 1-kb upstream UTR and EnhG, which are regions reported as enhancers but are overlapping gene bodies [23] (Fig. 3a) and (Additional file 16: Tables S15 and Additional file 17: Table S16).
Taken together our results agree with the findings of pre- vious studies on sheep domestication, which concluded that the major differences between domestic and wild sheep genomes concern functional elements close to genes rather than intergenic or distal enhancers [18].
Site frequency analysis
To complement our scan for selective sweeps in 20-kb bins and to exploit all the information from whole- genome sequences, we studied the differences in allele frequencies between European taurine and Asian indi- cine populations for 23,494,872 SNPs and between Afri- can taurine and indicine for 22,943,179 SNPs (Fig. 3a) and (Additional file 18: Tables S17 and Additional file 19:
Table S18). We found that only a small proportion of each set of SNPs, i.e. 228,908 (0.97%) and 26,561 (0.11%) SNPs, respectively, presented a ΔAF higher than 0.9, which indi- cates that they are close to fixation between the European taurine and Asian indicine populations, and between the African taurine and indicine populations, respectively.
Given the high level of divergence and the comparatively low correlation of allele frequencies between taurine and indicine cattle (Fig. 1) and (Fig. 3a) and (see Additional Table 2 Top 10 enriched terms from Mouse Genome Informatics (MGI) phenotype in GREAT for the identified Bos taurus selective sweeps in the comparison Bos indicus versus Bos taurus
242 20‑kb windows p‑ value < 0.05
Term name Binomial raw
p‑value Binomial test
FDR Q‑value Binomial fold enrichment
Binomial observed region hits
Binomial region set coverage
Genes
Increased exploration in
new environment 2.08 × 10−14 1.65 × 10−10 7.11 26 0.04 DRD3, FMR1, GRIA2, NPAS3, OPHN1, SH3KBP1
Decreased aggression 5.73 × 10−11 1.13 × 10−07 5.40 24 0.04 ARX, ESR1, FMR1, GRIA2, MAP6, NDUFS4, OPHN1
Abnormal kidney inter‑
stitium morphology 1.25 × 10−08 1.10 × 10−05 4.87 20 0.03 AGTR1, COL4A3, KIF3A, NPHP3, PDGFRA, TNFRSF1B, TRPS1, XDH
Abnormal social investiga‑
tion 9.88 × 10−07 2.17 × 10−04 3.42 22 0.03 AVPR1A, EXT1, FMR1, GRIA4, LRRTM1,
MAGED1, MAP6, NBEA, NPAS3 Abnormal strial marginal
cell morphology 1.64 × 10−06 3.09 × 10−04 10.25 8 0.01 COL4A3, ESRRB, KIT, NDP, SLC12A2
Abnormal startle reflex 1.87 × 10−06 3.44 × 10−04 2.37 37 0.06 BRE, CTNNA2, DRD3, ESRRB, FMR1, GLRB, GPR98, GRIA4, MECOM, MRO, NDUFS4, NPAS3, PHYKPL, SLC12A2, SLITRK6, TNFRSF1B
Abnormal frontal bone
morphology 8.19 × 10−06 9.55 × 10−04 2.89 23 0.04 BMP4, DISP1, EFNB1, HDAC8, HHAT, KIF3A,
MSTN, NOG, PDGFRA, SATB2, SP3, WNT9A Abnormal lens induction 4.21 × 10−05 3.00 × 10−03 4.20 12 0.02 BMP4, GRIP1, MAB21L1, PAX6, SOX1 Abnormal pain threshold 5.29 × 10−05 3.57 × 10−03 2.03 37 0.06 ADAMTS5, AFF2, ARX, BAMBI, EDNRB, ESR1,
EXT1, FMR1, GABRR1, GNAQ, GRIA2, GRIA4, HTR1F, LMO7, MC1R, NDUFS4, OPRK1, TRPM3
Abnormal fear‑related
response 8.43 × 10−05 5.09 × 10−03 2.98 17 0.03 ARX, ESR1, EXT1, FMR1, GRIK2, MAP2, SLITRK1
(See figure on next page.)
Fig. 3 Genomic feature enrichment in selective sweeps. a The strength of enrichment for 20 genomic features in 72 indicine‑specific regions assessed by overlapping genomic regions [50]. b Intersection of the delta allele frequency (ΔAF) with functional annotations derived from the reference UMD3.1 bovine genome. The number of SNPs in ΔAF bins is indicated on the left, and the M‑value (log2‑fold change) of the relative frequencies of SNPs in each functional category (on the right). The black line shows the number of SNPs within each (ΔAF) bin. c Intersection of the delta allele frequency (ΔAF) with functional annotations from the predicted regulatory elements in the cattle genome [20] and publicly available experimental epigenetic marks [49] and Fr‑AgENCODE [48]
b
c
UTR CDS Intron
Intergenic 1 Kb downstream
1 Kb upstream
Enh
ATAC
FrAgENCODE
H3K27ac
(Villar et al. 2015)
H3K4me3
(Villar et al. 2015)
ProxTFEnhG DistalTF
EnhBiv All predicted Gene annotation Epigenetic
P-adj < 0.05 a
file 1: Figure S5), sorting the sites under selection from those that display high ΔAF due to drift, is a challenging issue. Functional enrichment analysis (Additional file 19:
Table S18) confirmed our previous analysis at the level of genomic bins (Fig. 3a), since we observed a clear enrich- ment for UTR, coding regions and proximal regions such as promoter regions identified by H3K4me3 analysis in cattle liver [49], and predicted enhancer genic regions (Fig. 3b, c). The same analysis in African cattle (Addi- tional file 1: Figure S9, Additional file 20: Table S19 and Additional file 21: Table S20) agreed with these results.
Fixed coding mutations in the HELB gene in indicine cattle Comparison of the European taurine and Asian indicine genomes showed that a small proportion of the variants assessed (926 loci or 0.004% of those tested) were fixed for different alleles (ΔAF = 1). Annotation of these 926 loci revealed that only nine of them were located in exons (Additional file 22: Table S21). We detected one synony- mous mutation on chromosome X at 143,768,373 bp in ENSBTAG00000048102 or OFD1Y and eight muta- tions, three missense and five non-synonymous that were located within the HELB gene (Fig. 4a–c) and (see Additional file 18: Table S17) on BTA5 (47,713,856–
47,751,469 bp), which is within the previously reported 430-kb selective sweep on this chromosome (47,670,001–
48,100,000) (Fig. 2c). HELB functions as an ATP-depend- ent DNA helicase that is involved in DNA damage response [84, 85] and facilitates the recovery of the cells from replication stress during the S phase [86]. Non-syn- onymous mutations in HELB have been associated with male and female reproductive traits in tropical cattle [66]
and with Xeroderma pigmentosum, complementation group B, a skin pigmentation disorder in humans lead- ing to solar hypersensitivity of the skin [87]. In addition, point mutations in the HELB coding sequence have been identified in murine cell lines with temperature-sensitive DNA replication (Fig. 4c) [88]. Taken together, the muta- tions in HELB could lead to a modification of its DNA damage response function to better cope with different cell stresses associated with indicine tropical environ- ments such as constant high temperatures and high levels of UV intensity.
In African cattle, only four fixed coding mutations were identified between taurine and indicine populations (see Additional file 19: Table S18): one missense mutation in the NR4A1 gene (BTA5:27,982,214), which encodes a fibroblast growth factor involved in ovarian function [89], and three synonymous mutations in the coding regions of ENSBTAT00000005937, DRP2 and ADCK2 (Additional file 23: Table S22). Previously reported muta- tions in HELB were shown to be present in both taurine
(N’ dama) and indicine African cattle (Additional file 19:
Table S18).
Confirmation that point mutations in HELB are specific to indicine cattle
To examine whether mutations in the HELB gene are indicine-specific in a wider collection of breeds, we estimated the allele frequency of the HELB coding variant with the highest SIFT effect, i.e. rs447470311 (BTA5:47,726,121) in all the individual whole-genome sequences retrieved from Run6 (March 2017) of the 1000 Bull Genomes Project [24], i.e. 2709 whole-genome sequences corresponding to 97 classified breed compo- sitions [25]. We observed that only 36 breeds presented the G allele of rs447470311 corresponding to 100% of the indicine breeds or indicine admixed (Fig. 4d) and (see Additional file 1: Figure S10 and Additional file 24:
Table S23). However, it should be noted that this allele was also found, at a lower frequency, in some European taurine breeds (Fig. 4d) and (see Additional file 1: Fig- ure S10 and Additional file 24: Table S23), mostly Italian breeds, such as Marchiginana, Chinanina, Piedmontese, or Anatolian breeds, which are all known to have a his- tory of indicine introgression [56, 58, 59].
Finally, by assessing samples of ancient DNA from the 1000 Bull Genomes Project (Run6), we found that allele G was present (with an allele frequency for G = 0.067) in 15 of the samples tested from animals dating back to the roman empire and medieval era [90, 91]. Based on this result, we inferred that the G allele has not persisted in the taurine lineage because either of genetic drift or neg- ative selection within the European taurine breeds. This allele is also found in several current Iranian admixed individuals (allele frequency for G = 0.39, n = 9) and in Yak individuals (allele frequency for G = 1, n = 2) [25]
(Fig. 4d) and (see Additional file 1: Figure S8 and Addi- tional file 24: Table S23).
Mutations in the HELB gene and the HMGA2‑CNVR segregate independently
The HELB gene is located in a 430-kb selective sweep on chromosome 5 (47,670,001–48,100,000 bp) and is fixed in indicine cattle but not in taurine cattle (Fig. 2c).
This region also includes ENSBTAG00000026993, GRIP1 and part of HMGA2 (Fig. 2b, c) and (see Addi- tional file 1: Figure S11) for ARS-UCD1.2 coordi- nates). The latter gene is of particular interest because a 6.2-kb CNV that spans a segment of HMGA2 intron 3 in Nellore (indicine) cattle is associated with navel score [53] (Fig. 2c). Thus, we investigated whether the entire selective sweep region, i.e. including HELB and HMGA2-CNVR, was in linkage disequilibrium
or segregated independently, since independent seg- regation would explain the contribution of individual genomic elements in the region to multiple production traits in cattle. Towards this aim, we mapped bovine predicted topologically domains (TAD) in this region [92]. TAD are indicative of regions that physically interact more frequently with each other than with
other sequences outside of the TAD [93]. We found that two predicted TAD were located in the 430-kb selective sweep: one spanning, GRIP1, HELB, IRAK3 and ENSBTAG00000026993; and a second spanning HMGA2 (Fig. 2c), which strongly suggests that HELB and HMGA2 are located in two independent regulatory entities and segregate in an independent manner. Next,
hg19 chr12 : 66,696,335 - 66,737,423 ; bosTau6 chr5: 47,713,179 - 47,751,456
Bos Indicus Bos Taurus Human Chimp Gorilla Rhesus Mouse RatRabbit PikaPig Alpaca Camel Dolphin Sheep GoatCat DogFerret Panda Elephant Tas Devil Platypus
LL
LL LL LL LV LL LL LL LL LL LL V
LL LL LL LL LL LL LL LL LL LL AT-
PP PP PP PP PP PP PP PP PP PP --T
EE EE GA_GA_
ED ED DE EE QQ EE QQ E
NN NN ______
NS NS SN NN DN TN DP K
II II ______
TA TA TT TT TT AN T--- G SS SS ______
SS GS SS SS SS SS SS P
GG GG ______
KK ND NN NN PH PP EN G
SS SS ______
SS GS SS SS SS SS SS S
QQ QQ ______
QQ QQ QQ QQ QQ QQ QL Q
QQ QQ ______
KQ QQ QQ QQ EE EE QQ R
NN NN __C__C ND NE EN NN SG SS NN D
NN NN KR SN DS SN NN NN NN NE E
DD DD DD DD EE EE EE ED EE EN S
LL LL QQ LL AL LL LL RR RR LL V
DD DN EE DD KK KK KK EE EE EE L
AA AT AA PP AA AA AA AA AA AG F
SS SS EN K--- SS SG SS SS SS NH R
SG GG GG D--- GG GG GG GG GG GI N LL P
P E E N
N T
T F F N
N S
S P Q R
R N
N N
N E
E K
L K
L A
A G
G G S
b a
c
d
Fig. 4 Candidate SNPs in the HELB gene. a Representation of the HELB gene based on human hg19 coordinates (chr12: 66,696,335–66,737,423) and bovine bosTau6 coordinates (chr5:47,713,179–47,751,456). b Multiple sequence alignment of amino acids across representative mammalian species. In red: the candidate indicine‑specific mutations. c HELB amino acid sequence with the indicine‑specific mutations are highlighted in red.
Non‑synonymous mutations with substitution residues are located at positions 10, 788 and 791. In purple, residue 428, which had been previously associated to temperature sensitive murine cell lines [88]. d Allele frequency of the SNP rs447470311 located in chr5:47726121 across the 36 breeds in Run6 1000 Bull Genomes Project which contain the G variant [24]
to confirm that the mutations in HELB and the HMGA- CNVR segregate independently, we genotyped by whole-genome sequencing 71 animals from commercial breeds including 10 Brahman cattle, 5 Africander and 56 tropical composite, and assessed their genotype for rs447470311 in the HELB gene and for HMGA2-CNVR (Fig. 5a). Our results show that Brahman cattle (100%
indicine) are homozygous for the alternative allele (‘homozygous alternative’) of SNP rs447470311 and carry two copies of the HMGA2-CNVR (Fig. 5b) and (see Additional file 25: Table S24), whereas admixed or tropical composite animals displayed different combi- nations of genotypes at these two loci (Fig. 5b). Among the composite animals, all those that are homozygous for the reference allele at rs447470311, do not carry the HMGA2-CNVR. In contrast, all the animals that are homozygous for the alternative allele at rs447470311
carried one tandem repeat HMGA2-CNVR. It should be noted that, in our dataset, all Brahman cattle that were homologous at the rs447470311 alternative geno- type carried two HMGA2-CNVR (Fig. 5b). Finally, ani- mals that were heterozygous at rs447470311 carried either one HMGA2-CNV or no CNV. Thus, our results demonstrate that the genotype at the rs447470311 SNP in HELB and the HMGA2-CNVR segregate indepen- dently in admixed populations.
Discussion
The marked phenotypic, physiological and behavioural differences between taurine and indicine cattle offer the opportunity to identify which genomic loci and genes shape these fundamental differences. In this study, our aim was to exploit the population history of the tropi- cal beef cattle raised in Australia, which are a mixture
48,074,000 bp 48,075,000 bp 48,076,000 bp 48,077,000 bp 48,078,000 bp 48,079,000 bp 48,080,000 bp 7,511 bp
HMGA2
chr5: 48,073,547 - 48,081,058 HMGA2-CNVR
Gene
HMGA2 48,073,000 bp
48,074,000 bp 48,075,000 bp 48,076,000 bp 48,077,000 bp 48,078,000 bp 48,079,000 bp 48,080,000 bp 7,511 bp
HMGA2
chr5: 48,073,547 - 48,081,058 HMGA2-CNVR
Gene Sample2_CNV Sample1_No_CNV
Affected_Readsff Reads
Reads
Coverage
Coverage
[0 - 63]
[0 - 18]
Sample2_CNV Sample1_No_CNV
Affected_Readsff Reads
Reads
Coverage
Coverage
rs447470311 HMGA2-CNVR
Genotype 0 1 2
Composite 0/0 37 0 0
0/1 7 13 0
1/1 0 4 0
Brahman 1/1 0 0 10
a
b
Fig. 5 HMGA2‑CNV. a IGV screenshot of the HMGA2‑CNVR locus (in blue) and the visualization of the alignment of paired‑end reads. Sample1_No_
CNV shows the aligned reads of a composite animal presenting no CNV; Sample2_CNV shows the aligned reads of a Brahman animal presenting the duplication event. Paired‑end reads in green imply duplication with respect to the reference genome. These read pairs are oriented towards the outer sides of the predicted insert. Red colour reads indicate paired end reads with an insert size larger than expected (possible deletion). b Contingency table of SNP rs447470311 (chr5:47,726,121) versus HMGA2‑CNVR (chr5:48,074,233–48,080,443 (~ 6.2 kb) genotype
